AI: Timelines, Takeoff, and Risks with Michael Abramson

When will artificial intelligence surpass human capabilities, and what happens when an AI system begins to systematically design and improve itself?
In this session, Michael Abramson will lead an intellectual exploration into the competitive models, timelines, and existential questions surrounding advanced AI trajectories. Using contemporary quantitative forecasting frameworks (such as the benchmarks for Automated Coding and the transition to Artificial Superintelligence) as a macro backdrop, this discussion will explore the mechanics of the intelligence explosion without being bound to any single predictive model.
Our discussion will map out three high-level domains:

• Timelines and Critical Thresholds: We will examine the core milestones that forecasters track. This includes the pivotal transition where AI fully automates software engineering and AI research (often categorized as the Automated Coder milestone) and the subsequent horizon of Artificial Superintelligence.
• The Mechanics of Takeoff: What happens when the loop of AI development closes? We will analyze the theory of recursive self-improvement, exploring how a system capable of refining its own architecture could shift progress from a predictable, human-driven linear pace to a compressed, machine-driven takeoff.
• Systemic Risks and Governance: If timelines compress significantly, the window for human oversight shrinks. We will evaluate the planetary-scale risks, alignment challenges, and geopolitical pressures that emerge when technological acceleration outpaces traditional governance structures.

Whether you anticipate a slow, manageable transition or a rapid, non-linear takeoff, this session offers a structured sandbox to debate the variables, probabilities, and profound implications of the next epoch of intelligence.